2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_types.h"
24 #include "xfs_trans.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_inode_item.h"
39 #include "xfs_btree.h"
40 #include "xfs_ialloc.h"
41 #include "xfs_itable.h"
42 #include "xfs_dfrag.h"
43 #include "xfs_error.h"
45 #include "xfs_vnodeops.h"
48 * Syssgi interface for swapext
52 xfs_swapext_t __user *sxu)
55 xfs_inode_t *ip, *tip;
56 struct file *file, *target_file;
59 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
61 error = XFS_ERROR(ENOMEM);
65 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
66 error = XFS_ERROR(EFAULT);
70 /* Pull information for the target fd */
71 file = fget((int)sxp->sx_fdtarget);
73 error = XFS_ERROR(EINVAL);
77 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) {
78 error = XFS_ERROR(EBADF);
82 target_file = fget((int)sxp->sx_fdtmp);
84 error = XFS_ERROR(EINVAL);
88 if (!(target_file->f_mode & FMODE_WRITE) ||
89 (target_file->f_flags & O_APPEND)) {
90 error = XFS_ERROR(EBADF);
91 goto out_put_target_file;
94 ip = XFS_I(file->f_path.dentry->d_inode);
95 tip = XFS_I(target_file->f_path.dentry->d_inode);
97 if (ip->i_mount != tip->i_mount) {
98 error = XFS_ERROR(EINVAL);
99 goto out_put_target_file;
102 if (ip->i_ino == tip->i_ino) {
103 error = XFS_ERROR(EINVAL);
104 goto out_put_target_file;
107 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
108 error = XFS_ERROR(EIO);
109 goto out_put_target_file;
112 error = xfs_swap_extents(ip, tip, sxp);
119 kmem_free(sxp, sizeof(xfs_swapext_t));
133 xfs_bstat_t *sbp = &sxp->sx_stat;
134 bhv_vnode_t *vp, *tvp;
135 xfs_ifork_t *tempifp, *ifp, *tifp;
136 int ilf_fields, tilf_fields;
137 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
146 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
148 error = XFS_ERROR(ENOMEM);
156 /* Lock in i_ino order */
157 if (ip->i_ino < tip->i_ino) {
165 xfs_lock_inodes(ips, 2, 0, lock_flags);
168 /* Verify that both files have the same format */
169 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
170 error = XFS_ERROR(EINVAL);
174 /* Verify both files are either real-time or non-realtime */
175 if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
176 error = XFS_ERROR(EINVAL);
180 /* Should never get a local format */
181 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
182 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
183 error = XFS_ERROR(EINVAL);
187 if (VN_CACHED(tvp) != 0) {
188 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
189 error = xfs_flushinval_pages(tip, 0, -1,
195 /* Verify O_DIRECT for ftmp */
196 if (VN_CACHED(tvp) != 0) {
197 error = XFS_ERROR(EINVAL);
201 /* Verify all data are being swapped */
202 if (sxp->sx_offset != 0 ||
203 sxp->sx_length != ip->i_d.di_size ||
204 sxp->sx_length != tip->i_d.di_size) {
205 error = XFS_ERROR(EFAULT);
210 * If the target has extended attributes, the tmp file
211 * must also in order to ensure the correct data fork
214 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
215 error = XFS_ERROR(EINVAL);
220 * Compare the current change & modify times with that
221 * passed in. If they differ, we abort this swap.
222 * This is the mechanism used to ensure the calling
223 * process that the file was not changed out from
226 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
227 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
228 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
229 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
230 error = XFS_ERROR(EBUSY);
234 /* We need to fail if the file is memory mapped. Once we have tossed
235 * all existing pages, the page fault will have no option
236 * but to go to the filesystem for pages. By making the page fault call
237 * vop_read (or write in the case of autogrow) they block on the iolock
238 * until we have switched the extents.
241 error = XFS_ERROR(EBUSY);
245 xfs_iunlock(ip, XFS_ILOCK_EXCL);
246 xfs_iunlock(tip, XFS_ILOCK_EXCL);
249 * There is a race condition here since we gave up the
250 * ilock. However, the data fork will not change since
251 * we have the iolock (locked for truncation too) so we
252 * are safe. We don't really care if non-io related
256 xfs_tosspages(ip, 0, -1, FI_REMAPF);
258 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
259 if ((error = xfs_trans_reserve(tp, 0,
260 XFS_ICHANGE_LOG_RES(mp), 0,
262 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
263 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
264 xfs_trans_cancel(tp, 0);
268 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
271 * Count the number of extended attribute blocks
273 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
274 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
275 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
277 xfs_trans_cancel(tp, 0);
281 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
282 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
283 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
286 xfs_trans_cancel(tp, 0);
292 * Swap the data forks of the inodes
296 *tempifp = *ifp; /* struct copy */
297 *ifp = *tifp; /* struct copy */
298 *tifp = *tempifp; /* struct copy */
301 * Fix the on-disk inode values
303 tmp = (__uint64_t)ip->i_d.di_nblocks;
304 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
305 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
307 tmp = (__uint64_t) ip->i_d.di_nextents;
308 ip->i_d.di_nextents = tip->i_d.di_nextents;
309 tip->i_d.di_nextents = tmp;
311 tmp = (__uint64_t) ip->i_d.di_format;
312 ip->i_d.di_format = tip->i_d.di_format;
313 tip->i_d.di_format = tmp;
315 ilf_fields = XFS_ILOG_CORE;
317 switch(ip->i_d.di_format) {
318 case XFS_DINODE_FMT_EXTENTS:
319 /* If the extents fit in the inode, fix the
320 * pointer. Otherwise it's already NULL or
321 * pointing to the extent.
323 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
324 ifp->if_u1.if_extents =
325 ifp->if_u2.if_inline_ext;
327 ilf_fields |= XFS_ILOG_DEXT;
329 case XFS_DINODE_FMT_BTREE:
330 ilf_fields |= XFS_ILOG_DBROOT;
334 tilf_fields = XFS_ILOG_CORE;
336 switch(tip->i_d.di_format) {
337 case XFS_DINODE_FMT_EXTENTS:
338 /* If the extents fit in the inode, fix the
339 * pointer. Otherwise it's already NULL or
340 * pointing to the extent.
342 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
343 tifp->if_u1.if_extents =
344 tifp->if_u2.if_inline_ext;
346 tilf_fields |= XFS_ILOG_DEXT;
348 case XFS_DINODE_FMT_BTREE:
349 tilf_fields |= XFS_ILOG_DBROOT;
354 * Increment vnode ref counts since xfs_trans_commit &
355 * xfs_trans_cancel will both unlock the inodes and
356 * decrement the associated ref counts.
361 xfs_trans_ijoin(tp, ip, lock_flags);
362 xfs_trans_ijoin(tp, tip, lock_flags);
364 xfs_trans_log_inode(tp, ip, ilf_fields);
365 xfs_trans_log_inode(tp, tip, tilf_fields);
368 * If this is a synchronous mount, make sure that the
369 * transaction goes to disk before returning to the user.
371 if (mp->m_flags & XFS_MOUNT_WSYNC) {
372 xfs_trans_set_sync(tp);
375 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
380 xfs_iunlock(ip, lock_flags);
381 xfs_iunlock(tip, lock_flags);
384 kmem_free(tempifp, sizeof(xfs_ifork_t));